1. The Scope of the Invention
This invention relates to the synchronization of real-time data streams over asynchronous packet networks, in particular to the synchronization of timing and data over Ethernet and IP optical networks.
2. Backround
Industry experts recognize that Internet growth has created an unprecedented demand for additional core network capacity. The scalable and distributed nature of the Internet continues to contribute to it's growth on all of the fronts, including users, hosts, links and existing and emerging applications.
Internet users have been connecting at higher link speeds, and usage duration continues to grow, creating an exponential increase in traffic volumes.
Today's Metropolitan Area Networks (MAN) are based on SONET optical rings. The SONET protocol, originally designed for carrying voice, can no longer accommodate the requirements of a world that is becoming data-centric. The most suitable transport paradigm for the new packet-based traffic, increasing at a furious pace, is the Ethernet.
Though computing network environments have evolved and transmission rates have increased exponentially, the Ethernet network architecture has remained dominant. While a communication rate of 10 Mb/s was once considered as state-of-the-art technology, today an Ethernet local Area Network (LAN) may transmit at speeds of up to 10 Gbps, 20 Gbps and even more.
This is due to the development of optical fiber technologies that have enabled the transmission of digital data streams at rate of up to 10 gigabit/sec and more. This channel-to-channel technology involves the coupling of various computer systems together with optical fiber or with a fiber channel compatible electrically conductive (copper) cable, allowing data transmissions between machines separated by relatively great distances.
The existing switches and routers provide neither the performance nor the port density required to efficiently create a real-time simultaneous network with raw optical data.
One of the main characteristics of TDM (Time Division Multiplexing) networks is time synchronization (a natural requirement of it's time slots multiplexing method). In the Public Switched Telephone Network (PSTN) or in SONET/SDH networks, a clock master provides a primary time reference for synchronizing all of the network nodes (the time discipline). This master clock has an extremely long term accuracy of one part in 10−11. This reference time, the accuracy of which is called stratum 1, provides the reference clock to secondary network nodes with stratum 2 accuracy, and these in turn provide a time reference to stratum 3E and then stratum 3 nodes. This hierarchy of time synchronization is essential for the proper functioning of the TDM network as a whole.
A 10 gigabit optical network lacks the ability of transferring real-time synchronous data, since it is all based on a best effort from each of the routers within the Ethernet cloud and since the nature of pure packetized networks cannot ensure a stable or expected delay. This current situation does not allow good quality telephone voice transmissions to be created.
One of the main drawbacks that prevent optical networks from providing real-time data transmissions is the lack of synchronization between clock frequencies from the receiving and from the transmitting terminals connected to the optical network. In TDM applications, the transmitter and the receiver must share a common time base or at least be synchronized with one another in a Master/Slave chain, otherwise there cannot be a TDM connection.
This lack of synchronization results from several factors. The data Packets transmission in an asynchronous network suffers from random delays that are known as jitter and wander. The term “jitter” is used to describe short term signal variations, such as pulse position modulation frequencies that exceed frequencies of 10 Hz. The term “wander” is used to describe longer term variations of significant digital signal properties (e.g., zero level crossings) from their ideal positions in time and is applied to pulse position modulation frequencies below 10 Hz.
“Jitter” is typically attributed to additive Gaussian noise, whereas wander is typically attributed to slower varying environmental conditions.
Prior art methods and systems overcome the mentioned deficiencies by using a methodology called phase locked loop (PLL) (e.g. U.S. Pat. No. 6,246,738).
Another approach to overcoming this randomness when emulating TDM over an IP network is by using a buffer (FIFO) to smooth out all of the incoming data. This approach assumes that the proper time references are available. For the most part, however, the original time reference information is no longer available. The average time of emptying the buffer must be the same as the average rate of filling it up, otherwise we would be loosing data.
Another factor is known as packets slipping or the loss of data packets. This problem is more acute when relating to high rate data transfer networks such as an optical 10 gigabits network. Data packets arriving out of order result in substantial differences between the clock frequencies of the receiving device and transmitting device. In this case, trying to restore a clock frequency according to the incoming packet rate, by using it as a reference for the original transmission (+Jitter and −Wander caused by the network), would produce a frequency that is inaccurate, a result of the lost packets. Even the loss of only one packet out of a million packets creates a frequency error of 1 PPM (1×10−6), while the bit rate synchronization must satisfy 1×10−12 accuracy standards.
Prior art solutions, as proposed in U.S. Pat. No. 5,790,538, mainly address the problems of the actual loss of the data packets and methods for recovering them.
It was then suggested to provide time standards such as atomic clocks or GPS receivers to each edge terminal, thus relieving the IP network from the need to send and receive synchronization information. This suggestion does not provide any solution at all and would be a costly attempt. The incoming data has a certain momentary clock rate (according to the actual wander), which is influenced by routing interferences, temperature effects, network delays etc. If we try to extract that data by using an external clock, such as a GPS clock (even as accurate as an atomic standard), the data still arrives at a different average frequency. The slips and inaccuracy problems and loss of data remain the same. The local clock should represent the changing average of the incoming data at the time. This type of clock is called a “Breathing Networking Clock” that is capable of managing with the flexibility of the network, yet having a time discipline set to one central location, usually to the core switch that has the most accurate clock, and this accurate clock is also locked onto a better clock upstream.
There is also another solution based on retrieving the clock data from the nearest PSTN using existing linkages to the PBX via SONET. Such a solution requires the use of two competing networks and is based on the assumption that the PSTN is the same on both sides of the network. This is a very dangerous assumption, especially in a wide area IP network. But even if this assumption is right, the cost would be substantially high since the customer would have to pay for the use of two (and some times even four) different suppliers and networks.
BROADCOM HOMENETWORKING, INC introduces another solution in their international patent application, number WO 01/43325 A1. The BROADCOM application incorporates a synchronizing protocol for allowing voice phone calls over IP networks, which is specially adapted for the Home Phone line network alliance (HPNA). This synchronization solution between two end terminals is based on timestamp messages and network access expected jitter. Such a protocol methodology does not provide a complete solution for attenuating asynchronous network jitter/wander of up to a 250 microsecond amplitude. Thus, this solution cannot provide synchronization of real time data transmissions, and in particular the BROADCOM synchronization method does not comply with the requirements of the stratum 2/3/3e/4/4e environment, such as low output jitter (even in a case of 250 microsecond input jitter), hitless switch over between references, etc.
It is therefore a primary object of this invention to avoid the limitations of the prior art and provide real-time synchronous data transmission over asynchronous networks complying with the standards of Stratum 3/3e environment.